search

jik876 / hifi-gan

HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
MIT License
1.89k stars 500 forks source link

python environment #149

Open simonwindtner opened 11 months ago

simonwindtner commented 11 months ago

Hey, Can anybody suggest a python version for Mac M1 where all the requirements and versions are working correctly? with the proposed environment I get trouble with torch==1.4.0 as well as librosa and numba... Thanks in advance !

NathanJargon commented 11 months ago

You can just install a homebrew

/bin/bash -c "$(curl -fsSLhttps://raw.githubusercontent.com/Homebrew/install/master/install.sh)"

and install python for ARM

brew install python@3.9

and then start with a virtual environment 

python3.9 -m venv myenv
source myenv/bin/activate

dunno if you can download all package, but just do this and make use of ARM documentation

pip install torch torchvision torchaudio -f https://download.pytorch.org/whl/

chazarnik commented 1 month ago

Hey, Can anybody suggest a python version for Mac M1 where all the requirements and versions are working correctly? with the proposed environment I get trouble with torch==1.4.0 as well as librosa and numba... Thanks in advance !

Hi did you manage to solve it? I tried the same with windows and still the issue with numba persists due to llvmlite

simonwindtner commented 1 month ago

requirements.txt

@chazarnik this are my current package - I've made my on poetry package were everything is working, but in the meanwhile I run much more so - most of the packages maybe not necessary for just running HiFiGAN.

chazarnik commented 1 month ago

requirements.txt

@chazarnik this are my current package - I've made my on poetry package were everything is working, but in the meanwhile I run much more so - most of the packages maybe not necessary for just running HiFiGAN.

Thank you! I will try that too.

avdg-dev commented 1 month ago

@simonwindtner I am trying to run the code on google colab pro and I am facing the same issue where torch--1.4.0 is not supported. Does the requirements.txt file you have uploaded above work on google colab as well? if not, do you have any idea as to how i can solve this issue on google colab? Also, how did you train the model if i may ask? was it on your own gpu?

Thanks